Re: dask and python-xarray removed from testing
Antonio Valentino <[email protected]>
| Newsgroups | gmane.linux.debian.devel.python |
|---|---|
| Message-ID | <[email protected]> |
Il 24/08/26 08:46, Pieter Lenaerts ha scritto: > Antonio Valentino <[email protected]> schreef op 24 augustus 2026 08:08:51 CEST: >> Dear Diane, all, >> both dask and python-xarray have been removed from testing on July the 30th causing the removal of a relatively large stack of packages that rely on them. >> >> Both dask and, most of all, python-xarray are an important part of the modern Python scientific stack, so, IMHO, it is a real pity to have them out of testing for so long time. >> >> Apparently there is a quite complex (circular?) set of dependency involving at least dask, dask.distributed, python-sparse, python-xarray, python-pint, cfgrib. >> >> The blocking issue seems to be https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1143631, a single unit test of the dask test suite that goes in timeout on a single architecture (armhf). >> >> While te issue seems to be genuine it is a single, IMHO not fundamental, test on a single architecture, so, maybe, a temporary solution could be to just skip it to allow the migration. >> >> I could quickly prepare a patch if the solution is considered acceptable. >> >> Unfortunately, as a DM I do not have an immediate access to an armhf box. I could ask to get access to an armhf machine, but probably the process would take a little bit longer. >> >> For a longer term solution, IMHO, the first step would be to update the (rather old) dask version che is currently in the Debian archive. >> Some preliminary effort has been already started (see [1]), but probably more coordination is needed to complete the update. >> The last time that I looked into it, one of the main problematic point was the unavailability of pyarrow for Python 3.14 and for some of the currently supported platforms. I'm pretty sure that the situation is evolved since then. >> >> I would be more than willing to support the effort once we agree on the approach to follow. >> >> Please just let me know how to proceed. >> >> >> [1] https://lists.debian.org/debian-python/2026/04/msg00005.html >> >> kind regards >> -- >> Antonio Valentino >> > > Hi Antonio, > > I have been able to work on packages in different architectures using qemu. > > This may help here too. > > Ref my notes in <https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1123819#35> > > Checkout https://people.debian.org/~gio/dqib/ > > Install the right qemu-system : e.g. qemu-system-s390x in my case or other, like armhf in this case. > > Unzip the debian quick image backed zip. It may look weird. Hash for a filename. No extension. > > Read its README! It contains the line to start your qemu machine. > > Again, hope it helps. > > Pieter Thanks Pietet, Unfortunately I didn't manage to use your qemu images. The process stalls a tome point during the boot. The idea, by the way, was not to spend too much time trying to fix the issue in the current version of dask, but to implement a quick workaround to allow to package to migrate and focus on the update to a new version. Please note that the issue with the current version of dask on armhf could be in the test code or some limitation of the test environment rather then a real SW bug. kind regards